The accurate classification of causes of death is fundamental to progress in preventive medicine. Classifying death by poisoning may be specially difficult because the exact distinction between deliberate and accidental poisoning is inherently difficult for the coroner to make. Thus figures for suicidal poisoning are likely to be under-estimated because of ambiguous evidence and because of the stringent legal definition of suicide which coroners must apply. An examination of the drugs involved in poisoning cases, comparing their frequencies in the suicide and accident categories, may throw some light on the validity of coroners' present methods of classifying and provide evidence about the limits of error of under-estimation of the suicide rate.